Term project
(from 30 to 50 points)
You should choose and implement different projects to accumulate your grade to 40 points. You should prepare a video presentation along with your code for each project and submit it.
When you read the explanation there are blue points and red points.
From blue points, you must get at least 40 points, and from red points, you can get 20 extra points. However, you must
You have a deadline on 29th April 10:00 pm, and if you submit them earlier, you might get one chance to improve your projects and resubmit them before the end of the semester.
You cannot train anything from scratch. You are using an existing model and, if needed, refine them. To search for implemented models, you could search Github / Huggingface and use an existing model.
1- Create an application that makes the given image younger or older. You should test it with your photo. Again, you should experiment with quantization and pruning and compare the result, 10 points.
· By implementing this approach on the smartphone (Android and iOS), disconnected from the cloud and the Internet, you can earn 5 more points.
2- Train one of the open-source text-to-image models, which gets some of your images along with your name, e.g., AAAAA, and fine-tune the text-2-image model in a way that you can give a text prompt, and it generates a new image that includes your face, 10 points.
· By implementing this approach on the smartphone (Android and iOS), disconnected from the cloud and the Internet, you can earn 5 more points.
3- Use 3D Gaussian splatting and build a 3D model of yourself, capturing images from different angles. Please be sure to use a library for this purpose. 10 points.
· By creating a game character for yourself (3D Gaussian splatting/NeRF result) and being able to convert it into a 3D model, and deploy it in Unity or Blender, you can earn 10 more points.
4- Try to implement LLM agents (use small models for simplicity) that get a query, evaluate the results of each other, and present the highest quality answer; you need to study existing agentic LLM models and use one of the existing frameworks, 10 points.
· By incorporating a web search function (you can use duckduckgo search API or any other open search APIs) for your result construction, you can earn 5 more points.
5- Experiment with one audio voice replication model. I mean, give some audio samples to a model, and then it should create synthetic/fake audio using our given text, 10 points.
· By implementing this approach on the smartphone (Android and iOS), disconnected from the cloud and the Internet, you can earn 10 more points.
6- Find and deploy a model in which the user types a sentence about his/her mood, and it generates music based on the user’s mood, 10 points.
· By implementing this approach on the smartphone (Android and iOS), disconnected from the cloud and the Internet, you can earn 5 more points.